Application of Support Vector Machines to Approximate Objective and Constraint Functions of an Inverse Electromagnetic Problem

摘要

The Support Vector Machine (SVM), which is a state-of-the-art for linear and nonlinear input-output knowledge discovery, is proposed as a response surface model to approximate objective and constraint functions of an inverse problem. The detail formulations of the SVM based approximation model using ε-insensitive loss function are derived. Primary numerical results are reported to demonstrate the feasibility, performance, and robustness of the proposed SVM approximation technique in solving practical engineering design problems.